Integrated Structural-Electromagnetic Optimization Of Relfector Antennas Based On Kriging Interpolation

Reflector antennas are typical mechatronic systems. Along with theirdeveloping directions at large-diameter, high-frequency and fast-response, therequirements of their electromagnetic (EM) performances are getting increasinglystrict. In order to fulfill these requirements, numerical optimization methods shouldbe used on the basis of integrated structural-electromagnetic simulation. In thispaper, under the sport of a preliminary research project, this problem isinvestigated with focus on the techniques of transmitting the reflector distortiondata calculated by structural analysis software into the electromagnetic simulationtool. The main works are as follows.1. To observe and study the influence of the detailed distribution of errors ofthe reflector plate on the electromagnetic performances. Different distributions ofthe errors with the same rms(root-mean-square) are imitated by Gaussian MixtureModels(GMM). The EM performances are simulated respectively and some usefulconclusions are obtained.2. In order to transmit the distortion information of the reflector into the EMsimulation software, Kriging model is used to interpolate the axial displacementsof the reflector. The axial displacements at the nodes of the electromagneticanalysis meshes can be easily calculated by the Kriging model. Thus, the structuraland electromagnetic simulation software can be linked by the Kriging model, andthe integrated simulation is implemented.3. An integrated structural-electromagnetic optimization model for parabolicantennas is constructed. A surrogate-model-based optimization strategy is used toeffectively solve the optimization problem. Slack variables are introduced toprevent fail of the optimization procedure when there exits no initial feasible point.Compared with traditional gradient-based optimization approaches, the number ofreanalysis needed is reduced considerably. As an example, the backup structure ofan8m antenna is optimized. The results demonstrate the effectiveness and benefitsof the proposed models and methods.